AISOSolution for Sovereign AI

TPIsoftware and PHISON aiDAPTIV: Deploy On-premises AI at Speed and Scale

Consult an Expert

Build AI With Ease

AISO Portal is where you find all-in-one solutions for your business

Unlimited Tokens

Fully leverage AI capabilities with unlimited token usage for accelerated innovation

Ready to Use

Hardware and software that align with business needs without technical complexity

Enhanced Compatibility

Data remains on premises with reduced risks of security and incompatibility

Regulatory Compliance

AI infrastructure for highly regulated industries such as finance, healthcare, manufacturing and government

More on AISO Portal

Unlock Computing Capabilities with PHISON aiDAPTIV

Training Time (hours) by Model Size Single node, 4× GPU – aiDAPTIV™ vs. GPU-only
with aiDAPTIV™ GPU card only
Training set size and minimum GPU count by model size
Training Set Size 140 GB 260 GB 1,400 GB
Min GPU Count aiDAPTIV™ / GPU-only 4 / 4 4 / 6 4 / 30
Scaling is linear based on GPU count and model size. Training based on 10M tokens.

Boost Processing Capacity with Efficiency and Performance

In a single node 4 GPU setting, GPU with aiDAPTIV is 4 times faster in running a 70-billion parameter model compared with GPU only, supporting models that are 10 times larger than DELL and HPE
Quicker Time to First Token for Faster Inference aiDAPTIV™ vs. GPU card only – 8B model · hover dots for details
Time to first token by token length aiDAPTIV remains below three seconds at 32K tokens while the GPU-only series fails after 2K tokens and reaches more than eight seconds. 9s 8s 7s 6s 5s 4s 3s 2s 1s 0 TTFT (s) 8 16 32 64 128 256 512 1K 2K 4K 8K 16K 32K Token Length Reduce TTFT by 25×
with aiDAPTIV™ GPU card only

Speedy Inference with aiDAPTIV Reduced Time to First Token by 25x

Inference workload and latency increases as training data gets a grow. OrientAI when equipped with aiDAPTIV, delivers ultra-fast inference in generating outputs. In the setup with the same compute resource, the model's inference time is reduced from 52 seconds to 2 seconds.

AISOIntegrated Solutions for On-premises AI

All at Once in One Place

For Enterprise

Built for large organizations who seek high-concurrency inference and multi-tenant governance

Workstation

Processor / Platform

  • NVIDIA 6000ADA
  • NVIDIA RTX Pro 6000 Max-Q W/S Edition

Feature

  • Max GPU x 2
  • PCIe Gen5 x 16
  • 2TB Memory

aiDAPTIV Specs

  • aiDAPTIVReady: AI100E 2TB-4TB x 2

Rack Server

Processor / Platform

  • NVIDIA A6000 / 6000ADA
  • NVIDIA RTX Pro 6000 Server Edition
  • NVIDIA HGX H200
  • AMD Instinct MI300X

Feature

  • Max GPU x 8
  • PCIe Gen5 x 16
  • CPU (Dual Socket)

aiDAPTIV Specs

  • aiDAPTIVReady: AI100E 2TB-4TB x 2

For Labs

Ideal for software developers who want to run a PoC or advanced AI models on edge devices

AI PC

Processor / Platform

  • Intel Panther Lake
  • AMD Ryzen AI Max

Feature

  • Max GPU x 1 • PCIe Gen4 x 4
  • Ultra-compact

Key Specs

  • aiDAPTIVReady：A100E 320GB-1TB M.2 2280
  • GPU: NVIDIA RTX 5060Ti / 5090 / AMD RX 9070

AI NB

Processor / Platform

  • Intel Panther Lake
  • AMD Ryzen AI Max

Feature

  • Max GPU x 1 • PCIe Gen4 x 4
  • Portable, ultra-compact

Key Specs

  • aiDAPTIVReady: AI20EH 1TB
  • GPU: Intel® Arc™ GPU B390

For Personal or Small Team

Designed for highly-efficient, low-power AI workloads in edge computing

miniPC

Processor / Platform

  • Intel ARC 140T GPU 16GB

Feature

  • Max GPU x 1
  • Ultra-compact

Key Specs

  • 32GB DRAM

AI BOX

Processor / Platform

  • NVIDIA Jetson AGX Orin
  • AMD Ryzen AI Max
  • NVIDIA GB10 (Grace Blackwell)
  • Intel Core Ultra Edge

Feature

  • Max GPU x 1
  • PCIe Gen4 x 8
  • Fanless

Key Specs

  • GB10 / AMD Ryzen AI Max (Unified memory of 128GB)

AISO: Where Goals Meet Reality

with AISO integrated solutions, the procurement and implementation process is streamlined, making it faster to operationalize AI with efficiency and governance.

AISO Portal